The prompt
Paste this into Claude, ChatGPT, Cursor, or any MCP-compatible client connected to your Bourd account. Replace the highlighted placeholders with your own values before sending.
I've attached a Search Console export of queries my domain ranks for. I want to use it to seed a starter prompt set in the {{ workspace name }} workspace on Bourd.
Search queries don't translate directly into AI prompts. People phrase questions differently in a chat than in a search box. Take a look at the export and propose how you'd cluster the queries by intent, then show me how a handful of them would read as prompts.
We can refine the clusters and rewrites before you create anything in Bourd. Once I'm happy, go and create the prompts with appropriate tags. 1 placeholder to fill in. The agent will call: list_workspaces, create_tag, create_prompt, and 1 more.
What this workflow does
Reads a Search Console export you paste or attach in chat. Groups queries by intent: branded, comparison, and category-entry. Rewrites each cluster into natural-language prompts suited to AI chat, not SERPs. Creates a tag per cluster, writes the prompts, and applies the tags. You end the conversation with a tagged set ready to run.
Why it works
Search queries and AI prompts are not the same shape. "best crm for agencies" is a search query. "What's the best CRM for a small digital agency managing 20 clients?" is the prompt version. The agent does the translation in bulk, then writes straight into Bourd, so the taxonomy is consistent from day one.
Why per-workspace, and why quarterly
The first run is the obvious case: a fresh workspace, no prompt list, no demand signal grounding the choices. The recurring case is the more useful one. Every quarter your Search Console export looks different. New queries appear that you started ranking for. Old queries lose volume. Categories you ignored last quarter become the cluster worth investing in. Re-running on the latest export catches drift the dashboard would not. Ask the agent to compare new clusters against your existing tagged prompts and propose which to keep, which to retire, and which gaps to fill.
Common failure modes
Three patterns to watch for. The agent rewrites queries it should leave alone. Branded queries like "[your product] pricing" do not need translation, so tell the agent to leave brand queries verbatim and only rewrite category and comparison queries. The agent invents prompts that were not in the export, padding a cluster with a question it thinks fits the theme even though the underlying query never appeared in your data. Tell it to cite the source query for every generated prompt. The agent over-clusters into 15 themes. Six is the right ceiling for a starter set. More clusters than that and the tags become noise rather than navigation.
What a good starter set looks like
About 20 to 30 prompts split across three to five tags. Branded prompts cover roughly a third of the set. Category-entry prompts cover another third. Comparison and decision-stage prompts cover the rest. Coverage matters more than breadth. If the agent comes back with one prompt per tag, that tag is too thin to navigate by. Fold it back in. Ask for a minimum of three prompts per tag and roll the rest into "uncategorized" until you have enough volume to spin out a new cluster.
How to extend it
Pair with an Ahrefs or Semrush export to add competitor-ranking queries you don't rank for yet. Ask the agent to flag any cluster with fewer than five prompts as under-covered. Finish the conversation with "run the branded tag against GPT-5 and Claude so I have baseline data by morning."
Tools the agent calls
The agent picks from these 4 Bourd MCP tools based on the prompt. You do not call them directly.
list_workspaces create_tag create_prompt tag_prompts
Pairs with
MCP agents chain across servers. Add these alongside Bourd and the same prompt can hand output straight to the tools you already work in.
Google Search Console Ahrefs Semrush